Personalized Jargon Identification for Enhanced Interdisciplinary Communication.

Area of Science:

• Computer Science
• Computational Linguistics
• Information Science

Background:

• Scientific jargon presents a barrier to cross-domain research.
• Existing jargon identification methods lack personalization, failing to account for individual researcher backgrounds.
• This gap hinders efficient knowledge discovery and interdisciplinary collaboration.

Purpose of the Study:

• To develop and evaluate methods for identifying scientific jargon tailored to individual researchers.
• To investigate the impact of researcher-specific knowledge on jargon familiarity prediction.
• To explore the utility of personal data in enhancing jargon identification systems.

Main Methods:

• Collected over 10,000 term familiarity annotations from 11 computer science researchers.
• Analyzed jargon familiarity variations across individuals, even within the same sub-domain (e.g., Natural Language Processing).
• Compared supervised and prompt-based machine learning approaches using domain, subdomain, and individual knowledge features.

Main Results:

• Jargon familiarity and information needs significantly vary among researchers.
• Prompt-based methods incorporating individual researcher data (publications, research subfields) achieved higher accuracy.
• Supervised methods demonstrated lower false positive rates, indicating a trade-off between accuracy and precision.

Conclusions:

• Personalized data integration is crucial for effective scientific jargon identification.
• Future systems should leverage individual researcher profiles for improved jargon translation and accessibility.
• This work advances the novel task of personalized scientific term familiarity prediction.